Post on 26-Feb-2023
Part III
ISSUES IN ONTOLOGY
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 153 19.10.2005 3:15pm
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 154 19.10.2005 3:15pm
7. Inexpressible Properties andPropositions
Thomas Hofweber
1. the relevance of inexpressible properties
for the large-scale debate about (talk
about) properties
Everyone working on metaphysical questions about properties or pro-
positions knows the reaction that many non-philosophers, even non-
metaphysicians, have to such questions. Even though they agree that
Fido is a dog and thus has the property (or feature or characteristic) of
being a dog, it seems weird, suspicious, or confused to them to now ask
what that thing, the property of being a dog, is. The same reservations do
not carry over to asking what this thing, Fido, is. There is a substantial and
legitimate project to find out more about Fido, but is there a similar
substantial and legitimate project to find out more about the property of
being a dog? Metaphysicians know that there is a straightforward way to
motivate such a project, and much of the contemporary debate in the
metaphysics of properties is in the ballpark of carrying it out. If we agree
that Fido has the property of being a dog, then there is something that is a
property and that Fido has. Thus we can ask about what this thing is that he
has. How does it relate to Fido? Is it concrete or abstract? Is it fully present
in each object that has it? And so on and so forth. Maybe the non-
philosophers are merely not used to asking such questions about unusual
entities such as properties, but they are equally legitimate for them as they
are for any other thing. However, even metaphysicians sometimes have
the nagging feeling that something has gone wrong in the metaphysics of
properties, and that a substantial metaphysical investigation into their
Thanks to John Perry, Sol Feferman, John Etchemendy, Johan van Benthem, MarkRichard, Kent Bach, Rich Thomason, Allan Gibbard, Jim Joyce, Eric Lormand, AdamMorton, Jessica Wilson, Peter Railton, and Ted Sider for their help. The beginnings ofthis chapter date back to the author’s attending ESSLLI 97 in Aix-en-Provence with supportfrom Johan van Benthem’s Spinoza Grant, which is hereby gratefully acknowledged.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 155 19.10.2005 3:15pm
nature is somehow based on a confusion. This is not based on a rejection of
metaphysics in general, but concerns the metaphysics of properties in
particular. One obvious way to defend that such investigations are based
on a confusion is either to reject talk about properties as being based on a
mistake of some kind, or to accept such talk, but hold that it is not literally
true. If property talk is not to be taken literally, then no wonder that an
investigation into the nature of properties is not on a par with an inves-
tigation into the nature of Fido. But these two options are not very
attractive. Property talk can’t just be rejected without paying a stiff
price, and construing it as fictional can only work if the fictional goes
way down to the basics of our ordinary discourse, a conclusion not many
are willing to accept.1 But there might be other options.
Let’s call a minimalist approach to the theory of properties a philo-
sophical theory that explains why substantial metaphysical projects into
the nature of properties are mistaken, while at the same time accepting
talk about properties, and construing such talk as literal and often
literally true. This can be contrasted with a substantial approach to
the theory of properties, which holds that substantial metaphysical
investigations into the nature of properties are not confused, but merely
unusual and maybe difficult. Basically all the work done these days in
the theory of properties is in the ballpark of a substantial approach to the
theory of properties, whereas a minimalist approach would be congenial
to the non-philosopher’s judgments. The hard question, however, is
how one could possibly defend such a minimalist approach. It would
have to explain what is wrong with the simple and straightforward
motivation of a substantial metaphysical theory of properties outlined
above. But what could possibly be mistaken about asking what things
these properties really are, given that one holds that there are proper-
ties? In this chapter we will outline some ways in which one can hold on
to a minimalist approach, both for properties and for propositions as
well. These minimalist approaches, however, will be threatened by
considerations about expressibility that rely on inexpressible properties
and propositions. Formulating such minimalist approaches and rejecting
the concerns about expressibility will be the main goals of this chapter.
To get a first idea about how a minimalist approach to the theory of
properties could go, we can note that talk about properties has something
in common with talk about truth, and this feature of talk about truth is
central for motivating minimalist theories of truth. To see this we merely
1 For someone who does accept it, see Yablo (2000).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 156 19.10.2005 3:15pm
156 | Thomas Hofweber
have to look at what would happen if we gave up talk about properties.
A substantial investigation into the nature of properties would be
mistaken if there are no properties at all, and if thus talk about properties
should be stopped. Not many people have actually endorsed this view, but
there are some who played with it. Quine, in ‘‘On What There Is’’ was one
of them.2 In a well-known passage Quine discusses a view that accepts that
even though there are red houses, red sunsets, and red roses, there is
nothing they have in common. To accept that there is something they
have in common would involve accepting quantification over properties or
universals. But to deny such quantification would allow for the view that
there are no properties such as redness, even though there are red things.
According to this view, quantification over properties should be avoided in
serious scientific talk, and should be understood as merely a loose and
popular way of speaking. Quine seemed to have thought that such a
rejection of quantification over properties was a feasible option, one that
would avoid the usual metaphysical quarrels. But quantification over
properties can’t be rejected without paying a stiff price. Quantification
over properties is not just something that we need when we formulate
metaphysical theories. It is something we need in ordinary, everyday
communication. Quantification over properties increases our expressive
power in a way that we rely on in everyday life, not just metaphysics. This
is quite analogous to the need we have for talk about truth in everyday life,
outside of metaphysical debates about the nature of truth. If we think that
what Jones said during the trial is nothing but the truth, but we can’t recall
what precisely he said we can only communicate our belief about Jones’s
trial performance by saying
(1) Everything Jones said during the trial is true.
If we knew what Jones said we could just restate it, case by case, and say
that Jones said it, and that this is all he said. But if we can’t remember we
need a truth predicate to communicate this. Similarly for quantification
over properties. Sometimes we need to use quantification over properties
to communicate what we want to communicate in everyday situations
that have nothing to do with metaphysical debates. I might know that
(2) There is a property Clinton and Reagan share, but Nixon doesn’t
have,
but not know what it is. If I knew I could just tell you what it is. I could
simply say
2 See Quine (1980).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 157 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 157
(3) Both Clinton and Reagan like sushi, but Nixon doesn’t.
But without knowing what it is I have to use quantification over
properties to say what I want to say. Quantification over properties,
just like a truth predicate, fulfills an expressive need we have quite
independently of debates in metaphysics. We need it in everyday life.
Even though talk about properties has a function in ordinary commu-
nication, it is not so clear what we do when we engage in it. In particular,
it is not so clear if one can’t accept quantification over properties but still
reject the substantial metaphysical questions about properties as being
based on a mistake. To make this clearer, we should distinguish two
general large-scale views about what we do when we quantify over, and
more generally talk about, properties. The first is that talk about prop-
erties is talk about some mind and language independent domain of
objects (or entities). These entities are out there, part of reality, inde-
pendently of our expressing them with predicates. When I say that there
is a property such that F then I say that there exists a certain entity
which is a property such that F. Let’s call this view about talk about
properties the externalist view. The externalist view takes properties to
be independent of and external to the language used to talk about them.
Properties are out there, waiting to be expressed by predicates, grouping
things together independently of any mind or language.
The second way of looking at talk about properties is that it is not talk
about some language-independent domain of objects, and maybe even
not talk about some domain of objects at all. Rather properties are mere
shadows of predicates, as the metaphor goes, and quantification over
them is a device that increases our expressive power in a certain purely
logical or metaphysically thin way: quantification over properties is
nothing but a generalization over the instances. According to this view
such quantified statements will be truth-conditionally equivalent to
infinite disjunctions or conjunctions of the instances. The expressive
power we get from adding quantification over properties to our language
is thus equivalent to a certain infinitary expansion of our language.
Thus when I say that there is a property such that F then what I say is
truth-conditionally equivalent to the infinite disjunction over all the
instances in my language. For example, when I utter (2) then my
utterance will be truth-conditionally equivalent to something like
(4) Clinton and Reagan are tall, but Nixon isn’t, or Clinton and
Reagan are slim, but Nixon isn’t, or . . .
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 158 19.10.2005 3:15pm
158 | Thomas Hofweber
where for every predicate in my language there is a disjunct involving
that predicate.3 In particular, I am not talking about a language-inde-
pendent domain of properties as objects (or entities). The function that
such talk has in ordinary communication is supposed to come from the
metaphysically thin and purely logical increased expressive power
alone. Let’s call this view of talk about properties the internalist view.
The internalist view is understood broadly here, to include both the
view that properties are not entities at all, and thus talk about properties
is not talk about some domain of entities (call this the strict internalist
view), and the view that talk about properties is talk about entities,
though these entities are not mind-and language-independent (call this
the loose internalist view). Both views hold that quantification over
properties is equivalent to disjunctions and conjunctions over all the
instances in our language. The difference between loose and strict
internalist will be apparent in their different understanding of quanti-
fier-free talk about properties. A loose internalist will believe in a
sentence such as
(5) Redness is a color,
the phrase ‘‘redness’’ refers to or denotes some entity. A strict intern-
alist will believe that this phrase is not a denoting expression. I would
like to sideline the differences between loose and strict internalists for
now, and focus on the question whether or not quantification over
properties could be truth-conditionally equivalent to infinite disjunc-
tions and conjunctions over all the instances in one’s own language. The
difference between these two versions of internalism will reappear
later.4 So far the characterization of internalism is only a simplistic
first approximation. In the next section we will spell out more of the
details, including the relationship between predicates, nominalized
predicates, and quantification over properties.
3 This is, of course, only a simplistic first step. The details of this will have to be spelledout by the defender of such a view. In the next section we will see more about how such anassignment of truth conditions can go that has this result.
4 Internalist views have surfaced in the literature as the view that quantification overproperties is substitutional quantification. See Schiffer (1987), for example, where hediscusses an internalist view of properties and propositions, though not under thatname. Schiffer (288 n. 6) briefly discusses inexpressible properties and propositions andsuggests that some quantifiers over propositions are neither objectual nor substitutional.See also his more recent (1994, 1996).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 159 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 159
If strict internalism is the correct view about talk about properties then a
minimalist approach to the theory of properties follows. Strict internalism
does not reject talk about properties, and it accepts such talk as being
literally true. However, no substantial theory into the nature of properties
is legitimate. Talk about properties is not talk about any entities at all.
Thus no substantial investigation into the nature of these entities makes
sense. Even though it is true that there is a property that Fido has, there is
nothing to be said about what this entity is that Fido has, since there is no
entity that he has. We will see more details on how internalism can
maintain this, but for now we can say that if internalism were true then
a minimalist approach to the theory of properties would be the natural
conclusion to draw. Whether or not internalism is true thus seems of great
interest for the question whether or not substantial theories of properties
are ultimately based on a mistake.
To decide whether or not the internalist or externalist view is correct
might seem to be a substantial and difficult issue. It might seem to be
quite parallel to the debate about minimalist theories of truth. Is the
function of talk about truth captured by this increased expressive
strength, and is truth thus more a logical than a metaphysical notion?
Or is the increased expressive strength we get from talk about truth
merely a side effect of some metaphysically more heavy-duty main
function it has? This is a difficult issue that has been widely discussed.
And similarly, it might seem that to settle the difference between
internalist and externalist views about talk about properties one will
have to look at a number of difficult and substantial issues. It seems that
one has to look at the semantic function of quantifier phrases in certain
uses, issues about the relationship between quantification and ontology,
the function of talk about properties in ordinary communication, the
role of properties in metaphysics, or at some other similarly complex
issue. However, internalism can apparently be rejected for a much more
direct and simple reason. There is a simple argument that seems to show
that internalism is false, and that externalism is the only viable option.
This argument has nothing to do with the semantics of quantifier
phrases, the function of talk about properties, or the role of properties
in metaphysics. It is simply the argument that the internalist view
makes a too close connection between predicates and properties. The
internalist view seems to be committed to the claim that every property
is expressible by a predicate in our language. But that doesn’t seem to be
right. There are properties that are not expressible in our language, and
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 160 19.10.2005 3:15pm
160 | Thomas Hofweber
this is sometimes required for what we say to be true. Sometimes we say
that there is a property such that F, but the only properties that are F
are properties that are not expressible in present-day English. Therefore
the truth-conditions of such statements can’t be equivalent to infinite
disjunctions or conjunctions of sentences of present-day English where
a predicate has to be a witness for such a property. Therefore the
internalist view is wrong. The externalist view, however, has no such
problems. According to it, properties are language-independent entities.
Some of them might not be expressed by any predicate in present-day
English. It seems that the externalist view is the only viable option.
I will argue that considerations about inexpressible properties do not
refute internalism, but they do show something important about how
internalism should be understood more precisely. And once this is made
clear we will see that inexpressible properties are compatible with the
internalist view of talk about properties. In the first part of this chapter
I will show that this is so. I will show how it is consistent that there are
inexpressible properties and that quantification over properties is ex-
pressively equivalent to an infinitary expansion of our language without
such quantification (without any additions to the language). If I’m right
then internalism is not refuted by there being inexpressible properties.
After that I will argue that internalism also is not refuted by certain
other, more technical, arguments. This will make internalism a viable
option in the debate about the metaphysics of properties, and in the
debate about the function of our talk about properties. To decide be-
tween internalism and externalism is harder than it might initially
seem. Whether or not there are inexpressible properties is also an
important end-game consideration in a number of contemporary de-
bates, like the debate about minimalist theories of truth. The relevance
of the present consideration to this debate will be discussed after a more
refined version of internalism has been formulated. Finally, we will look
at how, why and in what sense different languages differ in their
expressive strength. Our discussion so far will suggest a hypothesis
about this, which we will formulate and discuss.
2. the formulation of internalism
In the above section we have roughly characterized internalism and its
rival externalism. For the following it will be necessary to give some
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 161 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 161
more details about how internalism is supposed to be understood more
precisely. In this section we will thus briefly discuss two important
issues about internalism that will clarify the position and contrast it
with externalism more clearly than we did above: what view about
quantification it is based on and how certain quantified statements are
associated with infinitary disjunctions and conjunctions.5
2.1. Quantification
If internalism is true then quantification over properties is equivalent to
infinitary conjunctions or disjunctions with the disjuncts or conjuncts
formed within one’s own language. This might seem like not much of a
substantial claim since it is often the case that quantified statements are
truth-conditionally equivalent to infinite disjunctions or conjunctions
formed within one’s own language. Take quantification over natural
numbers. When I say
(6) There is a natural number larger than 100,
then this is in fact equivalent to
(7) Either 1 is larger than 100 or 2 is larger than 100 or 3 is . . .
That such an equivalence obtains can be agreed upon by all sides in a
philosophical debate about numbers. What can be controversial, though,
is whether the equivalence obtains de facto or de jure. On the one side of
this distinction is the view that it just happens to be the case for natural
numbers that we have a term in our language for each one of them. So,
as it happens, a quantifier over natural numbers is equivalent to a
disjunction or conjunction, de facto. This, though has nothing to do
with quantification over numbers as such. Such quantifiers range over a
domain of entities, and it happens to be the case that we have a term for
each of them. On the other side is the view that quantification over
natural numbers has the function to generalize over the instances. It is
thus no accident that such quantifiers are equivalent to disjunctions and
conjunctions over the instances. That is their semantic function, and
such an equivalence thus holds de jure.
The same options apply to quantification over properties. The intern-
alist view is that such an equivalence does not merely hold as a matter of
5 I am indebted to Ted Sider and Dean Zimmerman for pressing me to make this clearer.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 162 19.10.2005 3:15pm
162 | Thomas Hofweber
fact, but that the function of quantification over properties is such that
this equivalence is guaranteed. It holds de jure, not merely de facto.
How this can be so will need to be explained, and it will require some
story about quantification. One rather implausible option that an
internalist has is to claim that all quantification is de jure connected to
infinitary conjunctions and disjunctions. This would be so, for example,
if all quantification is substitutional quantification. But this is not a very
plausible view about quantification in general. Some uses of quantifiers
clearly seem to have the function to make a claim about a domain of
objects out there, no matter how they relate to the terms we have in our
language to describe these objects. Some uses of quantifiers are not
equivalent to conjunctions and disjunctions formed within our own
language (such as quantification over real numbers, and many ordinary
everyday uses of quantifiers), some are equivalent to such conjunctions
and disjunctions de facto (such as quantification over natural numbers,
according to widely held views in the philosophy of mathematics, which
we will not challenge here), but according to the internalist, some such
equivalences hold de jure. But how can that be? How could it be that
quantifiers sometimes have a function such that this equivalence is
guaranteed, and sometimes apparently have a different function? An
internalist view about (talk about) properties or any other domain will
have to rely on a view about quantification that explains how this can be
so.
This chapter is not the place to discuss quantification in any detail,
and it is also not the place to provide a positive defense of internalism. In
this chapter we focus on whether or not internalism is refuted by
considerations about expressibility. I would like briefly to outline a
view about quantification that I believe to be true, and that I have
defended in Hofweber (2000 and 2005). This view about quantification
provides exactly what is required for internalism, about any domain, to
get off the ground. It does not guarantee that internalism about any
domain is true, but a brief look at it will be helpful for some of the
discussion below.
Many expressions in natural languages are semantically underspeci-
fied. That is to say, the language does not fully determine what contri-
bution an utterance of that expression makes to the truth conditions of
this utterance of which it is a part. There is a variety of different kinds of
such underspecification, from simple indexicals to subtle structural
underspecification. A good example of this is polysemy. The verb
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 163 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 163
‘‘run’’, for example, makes different contributions to the truth-condi-
tions to standard utterances of
(8) He ran the company well
and
(9) He ran the race well.
These different uses of ‘‘run’’ are not cases of ambiguity. We do not
have two different words that happen to be pronounced the same way.
Rather, it is the same word that can make different contributions to the
truth-conditions. These two uses of ‘‘run’’ are not unrelated, and it is no
accident that one and the same word is used in these two ways.
In Hofweber (2005 and 2000) I argue that ordinary natural language
quantifiers are semantically underspecified as well. One and the same
quantifier can make different contributions to the truth conditions, and
we can see from general needs we have for quantifiers in ordinary,
everyday communication that this is so and what these truth conditions
are. Natural language quantifiers have at least two readings, corre-
sponding to two general needs we have for them, and two functions
that they have. One is the domain conditions reading. When quantifiers
are used in this reading they make a claim about a domain of objects,
whatever it may be. The contribution to the truth conditions that
quantifiers make in this reading corresponds to the usual model theor-
etic semantics for quantifiers. We will also call this reading the external
reading. In addition quantifiers are used for their inferential role, in
their inferential role reading. In this reading they make a contribution
to the truth conditions that gives them a certain inferential role. In the
case of the particular quantifier, for example, it would simply be the
inferential role that ‘F(t)’ implies ‘Something is F’. This reading we will
also call the internal reading. That we have a need for quantifiers in
their internal reading, and that the internal and external reading of the
quantifiers do not coincide in truth-conditions is argued for in the two
papers mentioned. I’d like to add that the argument for an internal
reading of quantifiers is made not on metaphysical or ontological
grounds, but on the basis of general communicative needs in ordinary,
everyday communication. I will not repeat these arguments here.
The truth-conditions of quantified statements with the quantifier
used in its internal reading has to be such that a certain inferential
role results. The most direct way this is so is for the statement to be
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 164 19.10.2005 3:15pm
164 | Thomas Hofweber
equivalent to an infinitary conjunction or disjunction, formed within
one’s own language. Thus quantifiers in their internal reading are the
kind of quantifiers that an internalist about a certain domain will have to
rely on in their formulation of the internalist view. Using this view of
quantification one can state an internalist view, for example, that quan-
tifiers over properties are used in their internal reading. This will, of
course, have to be defended. How it is to be decided, even granting the
view about quantifiers outlined, whether or not ordinary uses of quan-
tifiers over properties, say, are in the internal or external reading is
quite another issue. The view about quantification outlined is, however,
quite relevant for our present discussion.
If it is true that quantifiers are sometimes used in an internal, inferential
role reading, and sometimes in an external, domain-conditions reading
then this has two consequences that are of philosophical importance. First,
internal quantifiers will in certain respects mirror Meinongian quantifiers,
without the ontology of non-existent objects. This point is discussed in
some length in Hofweber (2000). For example, a negative existential
statement such as
(10) The Fountain of Youth does not exist
will imply a quantified statement, with the quantifier used in its internal
reading, namely,
(11) Something does not exist, namely the Fountain of Youth.
So, in a sense we get quantification over non-existent objects, but we do not
get a domain of non-existent objects over which these quantifiers range, as
the Meinongians want to have it. The latter would only be so if (11) were
true with the quantifier used in its domain-conditions reading. But with
the quantifier used internally (11) can be literally true without there being
a domain of non-existent objects that the quantifier ranges over.
A second consequence of the two readings of the quantifiers is a
version of Carnap’s internal–external distinction about ontological
questions. I discuss this aspect in Hofweber (2005). If quantifiers are
semantically underspecified then there is more than one question that
can be asked with the words
(12) Are there properties?
According to the view about quantifiers outlined above, there are at least
two such questions, and this has a number things in common with the
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 165 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 165
position set out in Carnap’s (1956) essay ‘‘Empiricism, Semantics, and
Ontology’’. Just as Carnap thought, there will really be two questions that
can be asked with these words, and one of them is trivial, and the other one
is the one that is of interest to metaphysicians. But contrary to Carnap, the
questions that interest the metaphysicians are not meaningless. We will
thus get a version of a neo-Carnapian approach to ontology that affirms
ontology as a meaningful discipline, but holds on to a distinction between
internal and external questions about what there is.
A defense of internalism about properties and propositions is a task
that is much beyond the scope of this chapter. Here I attempt only to
investigate whether internalism about properties and propositions can
be easily refuted by considerations about expressibility. I hope to estab-
lish at the end that this is not so, and to do so it will be useful to have
some idea about what an internalist view will look like. There are a
variety of different ways one can spell out the background necessary to
formulate more precisely an internalist view, and I have outlined only
one of them, the one I prefer. Some of this background will be of use
later in our discussion, as will a better understanding how we assign
infinitary conjunctions and disjunctions to ordinary quantified state-
ments, which we will now turn to.
2.2. The Internalist Truth-Conditions
The internalist holds that quantification over properties and proposi-
tions is merely a logical device for increased expressive power that is
metaphysically innocent. Such quantification is supposed merely to
generalize over the instances, but not to range over a language-
independent domain of entities. And as we outlined above, this can be
understood as such quantification being truth-conditionally equivalent
to infinitary conjunctions or disjunctions. How quantified statements
get assigned these conjunctions or disjunctions has not been discussed so
far, and we will briefly do this here, as required for our discussion below.
To give a precise semantics for a fragment of a natural language is a
substantial and difficult task. It involves assigning expressions in that
fragment systematically to counterparts in a formal language with
clearly specified formal semantics. Such an assignment has to meet
certain conditions, such as preserving inferential relations. We will
not attempt to do this here for talk about properties and propositions,
nor do most other philosophical discussions of properties or propositions
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 166 19.10.2005 3:15pm
166 | Thomas Hofweber
do this. There are a number of difficult obstacles that need to be
overcome in doing this, no matter what ultimately one’s philosophical
views are. For propositions there is obviously the semantics of propos-
itional attitudes, but the difficulties in no way end there. To mention
one example, both properties and propositions exhibit the so-called
substitution failure. There is a clear difference between
(13) Lance feared that Jan will attack,
and
(14) Lance feared the proposition that Jan will attack,
as well as
(15) Being a philosopher is fun,
and
(16) The property of being a philosopher is fun.
How can we explain this if ‘‘being a philosopher’’ just stands for the
property of being a philosopher?6 We will not attempt to solve any of
the hard problems in this section, but merely outline how quantification
over properties and propositions can be assigned infinite disjunctions or
conjunctions as their truth conditions.
Let’s look at the simpler case of propositions first. We will take
recourse to the notion of a (grammatical) instantiation of a quantified
statement. So, for a statement such as
(17) He fears something I believe,
an instantiation is
(18) He fears that p and I believe that p.
To make this notion precise one will have to address a number of issues,
for example about the scope of quantifiers, a task not unique to the
internalist. Secondly, we take recourse to the notion of a (grammatical)
sentence. Finally, we assign to a quantified statement infinitary sentences
as follows:7 Suppose that ‘S[something]’ is a statement with a particular
6 See Bach (1997) and King (2002) for discussions of such examples for the case ofpropositions.
7 I will describe only the case of the particular quantifier. The universal quantifier isanalogous. I will not discuss in this chapter how to extend it to a treatment of generalized
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 167 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 167
quantifier over propositions, and ‘S[that p]’ is an instantiation of the
former. Let ‘Vp S[that p]’ stand for the disjunction over all instances of
‘S[that p]’, whereby for every sentence of our language there is one
instance of ‘S[that p]’, replacing ‘p’ with that sentence. Equivalently, for
present purposes, bind ‘p’ with a (particular) substitutional quantifier with
the substitution class being all the sentences in our language.
The case of quantification over properties is a little more complicated
since quantification over properties can apparently occur in subject as
well as in predicate position, and there is an issue about higher-order
predication. Whether or not there ever is quantification into a predicate
position is controversial. Possible examples are sentences such as
(19) He is something I am not (namely rich).
It is not clear whether this is quantification into predicate position since
the ‘is’ of the predicate is still present. In any case, an internalist can
specify the truth-conditions of such utterances quite directly, again by
taking recourse to the notion of an instantiation of such a quantifier, and
that of a predicate of our language. In this case an instance will be
something like
(20) He is F and I am not F.
The truth-conditions of the quantified statement are then simply the
disjunction over all the instances formed with predicates in our own
language. Thus in this case it would be ‘VP (He is P and I am not P)’.8
Quantification over properties, however, often is not of this kind.
Often the instances of quantifiers over properties do not directly involve
predicates, but expressions such ‘‘the property of being F’’ or ‘‘being F’’.
One example would be
(21) There is a mental property which is not a physical property
(namely the property of feeling pain).
Here the instances would be something like
quantifiers. This brings in further issues that are not central for our main discussion, inparticular the role of formal models in semantics.
8 Here the notation is an adoption of the above one we used for propositions above. ‘P’here is a predicate, whereas above ‘p’ was a sentence. The other changes for the case ofproperties are analogous.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 168 19.10.2005 3:15pm
168 | Thomas Hofweber
(22) Being F is a mental property and being F is not a physical
property.
We will have to expand our account to include such cases of quantifica-
tion over properties as well. This can be done quite directly by exploiting
the connection between a predicate, such as ‘is F’ and its nominalization
‘being F’ or ‘the property of being F’. In cases where a quantifier over
properties is a quantifier into a subject position we still form a conjunc-
tion (or disjunction) where there is a conjunct corresponding to each
predicate, but now the predicate appears in its nominalized form. Let
‘[nomF]’ stand for the nominalization of the predicate ‘F’. Then the
infinitary disjunction assigned to a quantified statement that quantifies
into subject position is the disjunction of all the instances such that there
is an instance for every predicate in its nominalized form. In the case of
‘Some property is G’ it is the disjunction ‘VP([nomP] is G)’.
One final issue to address briefly before we can return to our main
discussion is the issue of higher-order predication. Since properties
themselves can have properties there is a well-known division in the
theory of properties between those who take a typed and those who take a
type-free approach. For the former, all properties implicitly come with a
type, and every quantifier over properties only ranges over some type of
properties or other. In cases of higher-order predication properties of
higher type are predicated of properties of lower type. There are many
options one has in spelling this out in some more detail, and we will not
get into them here. For a type-free approach one denies that properties
come in types and claims that quantifiers over properties range over all of
them. In predication properties can be applied to all others, even to
themselves, in principle. To do this the predicate occurs nominalized in
the subject position, and regularly in the predicate position, as in
(23) Being a property is a property.
In our notation we can write self-application as
(24) [nomP] is P,
or
(25) P ([nomP]).
We will discuss this topic again later in the chapter, in connection with
the paradoxes. An internalist has both of these options available as well,
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 169 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 169
and the hard work that needs to be done in a typed approach, like
assigning types to particular occurrences of quantifiers, can be carried
over to a typed internalist approach as well. Instead of types being
assigned to properties directly they will be assigned to predicates, and
disjunctions and conjunctions will be formed involving only predicates
of the appropriate type. In a type-free version this will not be necessary.
We will return to this later.
With the internalist truth-conditions outlined above we can now see
in outline how an internalist would assign infinite disjunctions and
conjunctions to a large variety of quantified statements. For example,
(26) There is something we have in common
will by the above account be equivalent to the disjunction over the
instances formed within our language.9 In this case it is quantification
into subject position, and thus we get
(27) VP (we have [nomP] in common),
which is equivalent to
(28) VP (you have [nomP] and I have [nomP]),
which in turn, granting the equivalence between ‘x has the property of
being P’ and ‘x is P’, is equivalent to
(29) VP (you are P and I am P).
Whether or not a predicate occurs nominalized or regular in a disjunc-
tion is determined by the grammar of the sentences in which it occurs.
To avoid confusion, but at the price of extra notation, we will make this
explicit in the following, using the above notation.
Now it is time to return to our main discussion: is internalism easily
refuted by considerations about inexpressible properties? Since an
internalist claims that all such disjunctions are formed using predicates
from our own language, it would seem that internalism is refuted by
there being inexpressible properties. We will now look at this in detail.
9 One further issue that will have to be addressed is quantifier domain restriction. I daresay here that both the internalist and the externalist have the same or at least analogousoptions available, whether or not quantifier domain restriction is properly understood as asemantic phenomenon. A discussion of this and a survey of the options would take us toofar off course, however, so we will not attempt to do this here.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 170 19.10.2005 3:15pm
170 | Thomas Hofweber
3. internalism and the inductive argument
3.1. Inexpressible Properties and the Inductive Argument
I take it that we all believe that there are properties inexpressible in
English. It is, however, not so clear why we accept it. After all:
1. For a property to be inexpressible in a language means that no
predicate (however complex) expresses it. Simply because there is
no single word in a certain language for a certain property doesn’t
mean it isn’t expressible in that language.
2. We can’t be persuaded that there are properties inexpressible in
English by example. One can’t say in English without contradic-
tion that the property of being F isn’t expressible in English.
So, why again do we believe that there are inexpressible properties?
There are a number of different arguments for there being inexpress-
ible properties. We will look at several of them in this chapter. The
simplest and most important argument is the following:
Even though we can’t give an example of a property inexpressible in
English, we can give examples of properties not expressible in older,
apparently weaker languages. For example, the property of tasting
better than Diet Pepsi is not expressible in Ancient Greek. So, there
are properties expressible in English, but not in Ancient Greek. In
addition, we have no reason to believe that English is the final word
when it comes to expressing properties. We can expect that future
languages will have the same relation to English that English has to
Ancient Greek. Thus, we can expect that there are properties inex-
pressible in English, but expressible in future languages. In short,
there are properties not expressible in present-day English.
Let’s call this argument for inexpressible properties the inductive argu-
ment. It is a powerful argument. The main task for the next few pages
will be see whether or not the inductive argument refutes internalism
about talk about properties.
Internalism seems to be committed to a view that might be called
expressive chauvinism, the view that our present language is somehow
better than other languages when it comes to what can be expressed.
Our present language can express everything there is to express,
whereas other languages can’t express everything. But expressive chau-
vinism has to be rejected. Whatever reason we have to believe that our
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 171 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 171
language is expressively better than other languages, it will make it
plausible that some other (possibly future) language will be better in
that respect than our present language. Internalism has to free itself
from expressive chauvinism if it wants to be a contender in the debate
about properties.
3.2. Some Distinctions
What does it mean for a property to be expressible in English? Well,
that there is a predicate of English that expresses it. But that could
mean at least two things. On the one hand, it could mean that there is
a predicate of English that expresses this property in the language
English. On the other hand, it could mean that there is a predicate
of English such that a speaker of English expresses this property with
an utterance of that predicate. Which one of these we take will make a
difference for the issue under discussion. To illustrate the difference,
consider:
(30) being that guy’s brother.
This predicate does not express a property simpliciter, it only expresses
one on a particular occasion of an utterance of it, that is, in a particular
context. In different contexts of utterance it will express different
properties. However,
(31) being Fred’s brother
expresses a property independent of particular utterances, or better,
expresses the same one in each utterance.10 If ‘‘that guy’’ in an utter-
ance of (30) refers to Fred then this utterance of (30) will express the
same property as any utterance of (31) will. However, there might be
contexts in which an utterance of (30) will express a property that can’t
be expressed by an ‘‘eternal’’ predicate such as (31).
So, when we ask whether or not a property P is expressible in a
language l we could either ask
1. whether or not there is a predicate F (of l ) such that in every
context C, an utterance of F (by a speaker of l ) in C expresses P, or
10 I assume that ‘‘Fred’’ and ‘‘brother’’ are disambiguated, i.e. with respect to whetherwe talk about a monk or a sibling, and whether it’s Fred Dretske, Fred Astaire, FredFlintstone, or any other Fred.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 172 19.10.2005 3:15pm
172 | Thomas Hofweber
2. whether or not there is a predicate F (of l ) and a context C such
that an utterance of F (by a speaker of l ) in C expresses P.
Let’s call expressible in the first sense language expressible and express-
ible in the second sense loosely speaker expressible. The latter is called
loosely speaker expressible because it only requires for there to be a
context such that an utterance of F in that context by a speaker of l
would express P. Any context is allowed here, whether or not speakers of
that language actually ever are in such contexts. We can distinguish this
from what is factually speaker expressible. Here we allow only contexts
that speakers of that language actually are in.11 Let me illustrate.
Ancient Greek does not allow for the expression of the property
(32) tasting better than Diet Pepsi
in the sense of being language expressible. We can assume that. How-
ever, it seems that it is expressible in Ancient Greek in the sense of
being loosely speaker expressible. In the context where there is Diet
Pepsi right in front of a speaker of Ancient Greek he could simply utter
the Ancient Greek equivalent of
(33) tasting better than this
while demonstratively referring to Diet Pepsi. But since there was no
Diet Pepsi around during the time when Ancient Greek was a living
language, this context is not allowed when considering the question
whether or not this property is factually speaker expressible. In this
case, it seems that the property is not factually speaker expressible in
Ancient Greek, just as it is not language expressible in Ancient Greek.
Being language expressible implies being factually speaker express-
ible, which implies being loosely speaker expressible, and none of these
implications can be reversed (or so we can concede for now).
What all this shows is that both the inductive argument and the above
account of the internalist view of talk about properties was too simplis-
tic. In the latter it was simply assumed that contextual contributions to
11 To simplify, we consider someone only as a speaker of their native language. This isalso implicitly assumed in the inductive argument. We can also ignore complicated issuesabout identities of languages over time. For present purposes it does not matter what thedetails are about how long and under what conditions a language continues to be the same.Intermediate notions between factually speaker expressible and loosely speaker expressiblecan also be formulated depending on how strictly one takes ‘‘actually’’. This is of noconsequence for our discussion, though.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 173 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 173
content do not occur and that the truth conditions of talk about proper-
ties can simply be given by infinite disjunctions and conjunctions of
eternal sentences of the language in question. But that’s not always so.
Sometimes predicates express properties in some contexts that can’t be
expressed with eternal predicates. To say this is not to deny that
properties are shadows of predicates, just that they are shadows of
eternal predicates. Let’s call a version of internalism about talk about
properties extreme internalism if it claims that quantification over
properties is equivalent to infinite disjunctions and conjunctions formed
with eternal predicates. And let’s call a form of internalism moderate
internalism if it accommodates contextual contributions to content.
What we have seen so far is that extreme internalism can’t be right.
An internalist will have to endorse moderate internalism. But how is
this form of internalism supposed to be understood? How can internal-
ists accommodate contextual contributions to what is expressed by a
predicate while at the same time holding that quantification over prop-
erties is merely a logical device for increased expressive power?
3.3. The Problem
Here is the problem: even if an utterance of a sentence with demon-
stratives in Ancient Greek in the right context would express the
property of tasting better than Diet Pepsi, it is quite a different story
to extend this to an internalist account of the truth conditions of
quantification over properties. In fact, it seems that this can’t be done.
Let’s suppose, for the sake of the argument, that the only property of
beer that interests Fred is that it tastes better than Diet Pepsi. So
(34) There is a property of beer that interests Fred.
An internalist account of quantification over properties has to get this to
come out true. But it seems that this requires that there is a disjunct in
the infinite disjunction that corresponds to
(35) tasting better than Diet Pepsi.
This is no problem for our language, English. But if the disjunctions
have to be formed in Ancient Greek then it doesn’t seem to work. To be
sure, and as we have seen above, the property of tasting better than Diet
Pepsi is loosely speaker expressible in Ancient Greek. But how can this
be used in the infinite disjunction? After all, merely having
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 174 19.10.2005 3:15pm
174 | Thomas Hofweber
(36) tasting better than this
as part of one of the disjunctions won’t do unless the demonstrative
refers to Diet Pepsi. But how could it? The referent of a demonstrative is
fixed at least partly by the intentions of the speaker using it. And in an
utterance of (34) there are no such intentions that could back this up. For
one, one can utter (34) while having no idea what property it is that
interests Fred. And secondly, speakers of Ancient Greek will have no
idea what Diet Pepsi is, nor will they have any around to refer to
demonstratively. So, there is no way such speakers can fix the referent
of such a demonstrative to be Diet Pepsi. Thus the truth conditions of
quantified statements can’t be equivalent to infinite disjunctions and
conjunctions over the instances, even if the instances may contain
demonstratives. Thus it seems that internalism is refuted, after all,
even given the above distinctions.
3.4. The Solution
An internalist claims that quantification over properties is equivalent to
infinite disjunctions and conjunctions formulated using only the basic
vocabulary of the language on which the quantification occurs, plus
possibly extra logical tools. This would make quantification over prop-
erties merely a generalization over the instances. And it contrasts
internalism with externalism, which claims that such quantification
ranges over some mind-and language-independent domain of entities.
The above considerations suggest that extreme internalism should be
rejected. Extreme internalism is in trouble since not every object is
referred to with an eternal term, and thus what properties can be
expressed with eternal predicates is strictly less than what properties
are expressed with context sensitive predicates, namely in cases where a
demonstrative refers to an object that isn’t the referent of an eternal
term. These considerations show that what objects there are matters for
what properties and propositions there are, not merely what objects are
referred to with eternal terms. The case of demonstrative reference to an
object that isn’t the referent of any eternal term illustrates this. We
should not, however, give demonstrative reference too central a role in
this. What objects there are matters, not what objects can be referred to,
even with a demonstrative. So, if there are any objects that can’t be
referred to with a demonstrative, for whatever reason, these objects
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 175 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 175
would nonetheless be relevant to what properties there are. We thus
have to take ‘speaker expressible’ liberally here. Any object has to be
able to be contributed in a context. Context-sensitive expressions can
have terms in them that in a context can stand for an object. Demon-
strative reference is one way in which this can happen, but we will more
liberally consider the notion of a context contributing any object as the
value of a ‘demonstrative’ or context-dependent singular term. Thus
‘loosely speaker expressible’ has to be understood as expressible with a
predicate where context may contribute any object whatsoever as the
value of a ‘demonstrative’ or otherwise context-sensitive singular term.
This will properly accommodate the above insight that what objects
there are matters for what properties and propositions there are. So, an
internalist will have to claim that quantification over properties is a
generalization over all the instances of context-sensitive predicates,
with demonstratives being allowed to stand for any objects whatsoever,
but without requiring referential intentions or the like on the part of the
speakers. Let’s look at this more closely and with the help of artificial
languages to clarify the situation.
Let’s assume that the truth conditions of a fragment of a natural
language without quantification over properties is correctly modeled
with a certain formal language l . Adding quantification over properties
to that language should give us an infinitary expansion of l , according to
the internalist. Now, to accommodate demonstratives, we can do the
following. Add infinitely many new variables to l , which model the
demonstratives. Build up formulas as usual, but don’t allow ordinary
quantifiers to bind these new variables. To accommodate talk about
properties, we represent the truth-conditions of quantification over
properties with an infinite disjunction or conjunction as before, with
one difference. Whenever we form an infinite disjunction or conjunc-
tion we also existentially or universally (respectively) bind all these new
variables. Thus, now we do not simply represent ‘‘there is a property
such that F’’ as the infinite disjunction over all the instances
‘‘F([nomP])’’, i.e. as ‘‘VPF([nomP])’’. Now we take this disjunction and
add existential quantification on the outside binding all the new vari-
ables. So, we now represent ‘‘there is a property such that F’’ as
‘‘9y1, y2 . . . VPF([nomP])’’.
The new free variables play the role of the demonstratives in this
account, and the quantifier binding them plays the role of the arbitrary
contexts that we allow in loosely speaker expressibility. For example,
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 176 19.10.2005 3:15pm
176 | Thomas Hofweber
the disjunction that spells out the truth-conditions of (34) will contain a
disjunct corresponding to
(37) tasting better than yi.
Now there will be an existential quantifier that binds yi from the
outside. Since it will range over Diet Pepsis this disjunct will be true,
and thus the disjunction will be true. And this will be so independently
of there being a referring expression that refers to Diet Pepsi in the
language in question
However, there is no finite upper bound on how many of these new
variables will occur in these disjunctions. Since we allow, and have to
allow, every predicate to occur in the disjunction, we can’t give a finite
upper bound on how many variables can occur in these predicates. So, in
the infinite disjunction there will be infinitely many variables that have
to be bound, all at once, from the outside. But this can be done. We just
have to go to a higher infinitary logic. Not only do we need infinite
disjunctions and conjunctions, we need quantification over infinitely
many variables. Before we only used a small fragment of what is called
l v1,v, now we use a small fragment of l v1,v1. This latter logic also allows
for quantification over countably many variables.12 In both cases we
expand our original base language by only logical notions and no other
non-logical vocabulary.
Given this new model of talk about properties we have the following:
. Properties are shadows of predicates, but not shadows of eternal
predicates.
. Talk about properties gives rise to an infinitary expansion of the
original language, but not just to a small fragment of l v1,v, but to a
small fragment of l v1,v1.
So, the property of tasting better than Diet Pepsi is not expressible in
Ancient Greek in the sense of language expressible nor in the sense of
12 l v1,v is an infinitary logic that allows conjunctions and disjunctions over countablesets of formulas, but only quantification over finite sets of variables (as in regular first-orhigher-order logic). l v1,v1 allows for both conjunctions and disjunctions over countablesets of formulas, plus quantification over countable sets of variables. The basic language isusually the one of first-order logic, but one can define infinitary expansions of otherlanguages just as well. See Keisler (1971) and Barwise (1975) for much more on this. Inour case here we use only very small fragments of these logics. All these fragments will befinitely representable, for example, and smaller than the smallest fragments studied inBarwise (1975) or Keisler (1971).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 177 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 177
factually speaker expressible. It is however, expressible in Ancient
Greek in the sense of loosely speaker expressible. And by the inductive
argument we get that we have reason to believe that there are properties
that are not expressible in English, but we get that only when ‘express-
ible’ is understood in the sense of either language-expressible or fact-
ually speaker expressible. However, according to the present version of
internalism, quantification over properties has to be understood as being
based on what is loosely speaker expressible with predicates. Therefore
it will be true that
(38) There are properties that are not expressible in English,
if expressible is understood as being language expressible or factually
speaker expressible, but false, according to the internalist, if it is under-
stood as being loosely speaker expressible.
Internalism is not easily refuted with the inductive argument. Once
we distinguish between extreme and moderate internalism, and be-
tween different notions of expressibility, we can see that moderate
internalism is not refuted by the inductive argument. A moderate
internalist should endorse the inductive argument as showing some-
thing interesting and important about a difference in what is language
or factually speaker expressible in different languages. We will see more
about this in section 6 below. But before that we will have to have a
closer look at whether or not other arguments using considerations
about expressibility refute internalism. We can grant that extreme
internalism is refuted by the inductive argument, though I have argued
that moderate internalism isn’t refuted by it. We will now have to see
whether or not moderate internalism can be refuted as well, and thus
‘internalism’ will mean ‘moderate internalism’ from now on.
4. further arguments against internalism
using expressibility considerations
We have looked at the inductive argument above, and seen that endors-
ing it poses no threat to internalism about talk about properties, rightly
understood. The inductive argument, however, is not the only argu-
ment that tries to refute internalism using inexpressible properties. In
this section we will look at several other arguments for this conclusion,
and we will see that they are no problem for internalism either. To be
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 178 19.10.2005 3:15pm
178 | Thomas Hofweber
sure, I can’t claim to deal with all possible such arguments, but once we
see that the arguments discussed in the present section are no problem
for internalism it should be plausible that internalism is in fact not
threatened by issues about inexpressible properties. After that we will
look at some positive and more large-scale issues.
I will divide the further arguments against internalism from inex-
pressible properties into several groups: First, arguments that are modi-
fications of the inductive argument. Secondly, arguments that try to
establish that there are strictly more properties than expressible prop-
erties. These arguments thus try to establish that the cardinality of the
set of properties is larger than the cardinality of the set of expressible
properties. Thirdly, we will look at arguments that internalism is in
conflict with some lessons that have to be drawn from the semantic
paradoxes. After that we will look at an argument trying to establish
that we can ‘‘diagonalize out’’ of the expressible properties. Finally, we
will look at arguments that take recourse to modal considerations.
4.1. Modified Inductive Arguments
We have to see what reason one might have for believing that there are
properties that aren’t loosely speaker expressible. And, since this again
can’t be motivated by giving an example of such a property, one way to
go will be a version of the inductive argument, but this time an induct-
ive argument for there being properties that are not loosely speaker
expressible. And to start such an argument we have to point to a
property that is loosely speaker expressible in English, but not loosely
speaker expressible in, say, Ancient Greek. What could that be? The best
candidates for such properties are ones that relate to an area where there
is a substantial difference between Ancient Greece and us, such as
scientific understanding of the world. A tricky example is
(39) being a quark.
It might seem that it isn’t even loosely speaker expressible in Ancient
Greek. Whatever one’s prima facie intuitions about this are, we should
note that since this property is (language) expressible in present-day
English, but presumably not in English of ad, 1600 something must
have happened in the recent history of English that allowed for the
language expressibility of this property. So, how did we come to be able
to express it? That certainly is a hard question, related to some difficult
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 179 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 179
issues in the philosophy of science. Two possibilities come to mind,
though, namely:
� ‘‘being a quark’’ is a theoretical predicate of physics. It is at least in
part implicitly defined by the physical theory that uses it. Thus we
can express it because we have the theory.
� We can express the property of being a quark because we have been
in contact with observable phenomena that are caused by quarks,
such as effects they have on some measuring instrument.
If either one of these is the correct account then there is no problem for
the internalist. The reason is simply the following. If the first is correct
then the problem of expressing the property of being a quark reduces to
expressing the theory that implicitly defines ‘‘being a quark’’, plus
making the implicit definition explicit. Simply put, the property of
being a quark is the property of being such that the theory truly
describes you. Thus the problem is pushed back to the properties used
in the implicit definition of ‘‘being a quark’’, that is in the formulation
of the theory that implicitly defines it. In general, though, if the
apparent increased expressive power of new theoretical concepts comes
from their implicit definition in scientific theories (or from mixing those
with the above second point) then internalism is not in trouble.
If the second possibility is the right one then the increased expressive
power does come from being in contact with more objects. If we intro-
duced ‘‘being a quark’’ as
(40) being the kind of thing that causes these effects on the measur-
ing instrument,
or something along this line then being a quark is loosely speaker
expressible in older languages, though not language expressible. This
case thus essentially reduces to the case of the (standard) inductive
argument.
To be sure, these are only rough outlines of how this can work. How
such predicates work in general is very difficult to say. We should,
however, keep the fact in mind that something must have happened in
the last few hundred years that made the change from speakers of
English not being able to (language) express this property (at least not
with a simple predicate) to their being able to (language) express it with
just a few words. One easy explanation of how this might have hap-
pened is that speakers were able to express the property before, after all,
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 180 19.10.2005 3:15pm
180 | Thomas Hofweber
either with a complex eternal predicate, or with some non-eternal
predicate in the right context. If this is so then there is no puzzle how
we can now express it with a few simple words: we just introduced a
word to stand for a property that we could express already, though only
with a complex predicate, or only in special circumstances. But if this
isn’t so, what might have happened that made the difference? One
answer is holism, and it is hard to see what another answer might be.
We will get more into the details of this issue at the end of this chapter.
4.2. Cardinality Considerations
The second strategy to argue against internalism using considerations
about inexpressible properties takes recourse to cardinality consider-
ations. Such arguments try to establish that the set of all properties is
strictly larger than the set of all expressible properties. I would like to
divide these arguments into two groups. The first group takes recourse
to a principle that connects how many objects there are with how many
properties there are, and argues that since there are a certain number of
objects there are more properties than expressible properties. The sec-
ond group of arguments takes recourse to ‘‘closure’’ principles. These
arguments are based on that what properties there are is closed under
some general principle. Using such principles, the argument continues,
we can see that there are more properties then there are expressible
properties.13 Let’s look at these in turn.
4.2.1 Arguments from Objects
Here is a paradigmatic argument from objects against internalism:
Our language has only a finite base vocabulary, and only finite
combinations of it are allowed to form predicates that express
13 Arguments from closure principle don’t necessarily attempt to reach the conclusionthat the cardinality of the set of expressible properties is smaller than the cardinality of theset of properties. Some of them might try to establish only that the set of expressibleproperties is a proper subset of the set of all properties. In the relevant section below wewill mainly discuss closure principles that, if they were true, would lead to a cardinalityargument. Thus arguments using closure principles are dealt with under the heading ofcardinality arguments here. In section 4.4 below we will discuss additional arguments thatthe expressible properties form a proper subset of all the properties that are not cardinalityarguments.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 181 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 181
properties. Thus overall we can form countably many predicates. But
there are uncountably many properties. There are, for example,
uncountably many real numbers. And for every real number there
is the property of being larger than that real number, or other
properties of this kind. Thus there are uncountably many properties.
So, internalism has to be false.
This is prima facie a very plausible argument. But once we take into
account the distinctions that were drawn above we can quite easily see
that it is flawed. The argument would work against extreme internal-
ism, which holds that properties are shadows of eternal predicates. But,
of course, this is not the form of internalism we are discussing now.
Internalism has to be understood as moderate internalism, which holds
that properties are shadows of predicates, though not of eternal predi-
cates.14 The truth conditions of quantification over properties is under-
stood as being modeled by infinitary disjunctions as well as infinitary
first-order quantification. In particular, what is in the domain of the
first-order quantifiers will matter for the truth-conditions of quantifi-
cation over properties.
And once we consider this formulation of internalism we see that the
above argument provides no problem for it. If we grant, as is presup-
posed in the above argument, that real numbers are objects in the
domain of first-order quantification then it will be true according to
moderate internalism that
(41) For every real number there is a property that is the property of
being larger than that real number.
According to internalism the truth conditions of this sentence can be
spelled out as (in semi-formal notation):
(42) 8r9~vvW
P ([nomP(vi)] ¼ being larger than r)
And (42) is true, as can be seen as follows. Fix an arbitrary number r.
One of the disjuncts in the disjunction will be
(43) (being larger than vi ¼ being larger than r)
with a variable vi bound from the outside by an (infinitary) existential
quantifier. Since this variable ranges over real numbers, in particular
14 See the end of s. 3.2 for the distinction between extreme and moderate internalism.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 182 19.10.2005 3:15pm
182 | Thomas Hofweber
number r, there is a value to the variable that makes this disjunct true,
namely r. So, (42) is true.
Real numbers, as objects of the domain of first-order quantification,
can be the values of the variables that occur in the infinitary disjunc-
tions, which are bound from the outside by the (infinitary) existential
quantifiers. So, the more objects there are in the domain of first-order
quantification, the more properties are loosely expressible, and the
stronger is quantification over properties. The arguments from objects
thus rely on a version of internalism that is based on using eternal
predicates as the basis of expressibility. They do not affect the present
version of internalism.
4.2.2. Arguments from ‘‘Closure Principles’’
A second argument against internalism about properties is based on
considerations that properties are closed under certain principles. In the
simplest form these are principles of the kind that for any two properties
there is a property that is the conjunction of the two. So, if being a dog is
a property, and being a cat is a property, then being a dog and a cat is
also a property. This is, of course, a very innocent form of a closure
principle, but others are not so innocent. To argue against internalism
using closure principles one will have to find a closure principle C such
that
(i) we have good reason to believe that properties are closed under
principle C, and
(ii) properties being closed under C is incompatible with internalism
about properties.
We will in this section look at whether or not there are any good
arguments of this kind.
A principle that seems to meet condition (ii) is:
(P1) For every set S of properties there is a property of having all the
properties in S.
(P1) seems to be incompatible with internalism because of the argu-
ment:
Let’s agree that how many properties are expressible depends on what
objects there are in the domain of first-order quantification. So, if
there are k many objects in this domain, and if the base language is
countable then there will be k many properties that can be loosely
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 183 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 183
speaker expressed.15 However, there are more than k many proper-
ties. Since what properties there are is closed under principle (P1) and
since there are 2k many sets of loosely speaker expressible properties
there are atleast 2k many properties. For every set of loosely speaker
expressible properties there is a property of having all these proper-
ties, and for every such set this resulting property is different. Thus
there are at least 2k many properties.
This objection has several problems. First, but not most importantly, the
argument begs the question against the strict internalist, who thinks that
properties are not objects or entities, and are thus not available to be
collected into sets. Sets of properties can only be built if properties exist
as entities. But according to strict internalism, properties are not entities.16
Thus to talk about sets of properties is to assume that properties are
entities, which is one of the issues at stake here. Secondly, and more
importantly, if we allow sets of properties, and if we accept sets in general,
then these sets will be in the domain of the first-order quantifiers. In
particular, the assumption in the above argument that there are only k
many objects in the domain of first-order quantification is false if there are
also 2k many sets of properties, or sets of any kind. Thus the above
argument really is a version of an argument from objects, with the differ-
ence that it uses sets, rather than real numbers. Such arguments, we have
seen, should not bother the moderate internalist. Cardinality arguments
using sets of properties are thus no threat to moderate internalism.
One might attempt, though, to give a related argument that does not
rely on using sets. An opponent of internalism might argue that there
are certain principles we should accept that tell us that properties are
closed under certain operations, without ever taking recourse to sets of
properties. A simple example of this would be the claim that properties
are closed under conjunction. This can be formulated quite innocently as
(44) If P is a property and if Q is a property then P and Q is a
property.
Some principles like this, it seems, will have to be accepted by anyone,
and their acceptance does not beg any question against internalism or
15 k is an infinite cardinal number. If there are only finitely many objects then there willbe countably many expressible properties.
16 According to the other kind, loose internalism, properties are language-dependententities.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 184 19.10.2005 3:15pm
184 | Thomas Hofweber
externalism.17 Such principles take the form of a schema that claims that
if there are certain properties then there are also other properties. Can
we specify a schema of this kind that would refute internalism?
No. Every such schema that we can specify is compatible with intern-
alism. In fact, it is implied by internalism.18 Expressible properties are
closed under expressible closure conditions. Whatever the right-hand
side of a closure condition would say, it would give us a predicate for
expressing the relevant property. If the schematic letters stand for
expressible properties then we will have predicates that express them,
and the expressible closure condition will give us a recipe to construct a
new predicate that expressed the new property which the condition
claims there is. Thus any example of a schema that expresses a closure
condition is compatible with internalism.
I conclude that cardinality considerations do not refute internalism,
even though they seem on the face of it to be a serious problem for
internalism. Cardinality considerations would work against a simple
form of internalism, where only eternal predicates are allowed to express
properties, or where expressibility is understood as language expressi-
bility. Moderate internalism is not threatened by these problems.
4.3. Paradoxes
One quite different argument against internalism doesn’t claim that
internalism does not accommodate all the properties, but rather claims
that internalism allows for too many properties. This might be espe-
cially striking because of what we have seen in the above section. Are
properties really closed under all these expressible closure conditions?
There is, of course, property elitism, the metaphysical view that there
are only few and only very special properties, which disagrees with
this.19 According to one version of property elitism, for example, prop-
erties are not closed under disjunctions, and in fact there are no dis-
junctive properties at all. Such views, however, rely on heavy duty and
17 Some such principles might beg the question against some forms of property elitismto be discussed below.
18 Properly formulated, of course. In our case it would be the universal closure of
(62) If [nomP] is a property and if [nomQ] is a property then [ [nomP] and [nomQ] ] is aproperty.
19 The classic example is Armstrong in e.g. (1978, 1989), and others.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 185 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 185
controversial metaphysics. Simply because internalism isn’t compatible
with property elitism isn’t an argument against it. Internalism, as a view
about the metaphysics of properties, is naturally incompatible with
several other competing views about the metaphysics of properties.
There are, however, other arguments against internalism that argue
that internalism admits too many properties. These arguments are not
based on metaphysical considerations, but rather on the paradoxes.
What these arguments try to establish is that not every predicate
expresses a property. A simple argument of this kind is:
Even though we can’t give an example of a property that isn’t
expressed by any predicate (in our language), we can give an example
of a predicate that doesn’t express a property. It is the predicate:
(AP) does not apply to itself
If this predicate would express a property, P, then we can ask whether
or not P applies to itself, i.e. whether or not P(P) holds. And we can see
that P does apply to itself iff it does not apply to itself. Contradiction.
Thus there can be no property that is expressed by this predicate.
It is a not an uncommon reaction to conclude that this paradox shows
that there is no property that is expressed by (AP). In particular, there
are predicates that express no property, contrary to internalism. How-
ever, this reaction is premature.
The above argument relies on the fact that the lesson to draw from
this paradox is that there is no property expressed by this predicate. This
is problematic for several reasons. For one, it seems true to say that
there is a property that seems to lead to paradox, or that puzzled
logicians for decades, namely the property of not applying to oneself.
But more importantly, this account of denying that there is such a
property doesn’t really solve the paradox. The paradox can be formu-
lated in such a way that it doesn’t even take recourse to properties. Thus
trying to solve the paradoxes by adopting a certain view in ontology,
that certain properties do not exist, does not get at the heart of the
problem, and provides only an ad hoc way to block a certain formulation
of the paradox, but not a solution to the paradox. We can also formulate
the above paradox using only predicates, and not properties. Here is the
modified version:
Predicates are satisfied by objects. So, ‘‘is a dog’’ is satisfied by Fido.
Predicates can also be satisfied by predicates. For example, ‘‘is short’’
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 186 19.10.2005 3:15pm
186 | Thomas Hofweber
is satisfied by ‘‘is short’’. Now, call a predicate ‘‘heterological’’ if it
does not satisfy itself. So, ‘‘is short’’ does not satisfy ‘‘heterological’’.
Does ‘‘heterological’’ satisfy ‘‘heterological’’? By the usual reason-
ing: it does iff it doesn’t. Paradox.20
The above version of the paradox can’t be resolved by claiming that
certain properties don’t exist, or that certain predicates don’t express
properties, since we never took recourse to properties. In particular, no
one would conclude from this paradox that there is no such predicate as
‘‘heterological’’. So, if denying that there is such a property as not
applying to oneself doesn’t solve the paradox, but only blocks the
particular formulation of the paradox, then this doesn’t give us a reason
to believe that there is no such property. Denying that ‘‘doesn’t apply to
itself’’ expresses a property doesn’t solve the paradoxe. It can at most
require us to give the paradox a slightly different formulation.
We have seen that these paradoxes don’t require an internalist to
claim that some predicates don’t express a property. But an internalist
does not have to take this route. An internalist will have to spell out
anyway what will count as a predicate in making more precise what the
infinite disjunctions and conjunctions will look like. It is an option for an
internalist to spell this out in such a way that ‘‘does not apply to itself’’
is not a predicate. I don’t want to pursue this here, however.
How the paradoxes can be solved, if at all, is of course a completely
different story. We only need to note here that internalism and extern-
alism don’t seem to be importantly different in that respect. In particu-
lar, paradoxes are no more in conflict with internalism than they are in
conflict with anything else.
4.4. Diagonalization
One further more or less technical argument has to be dealt with. It uses
the technique of diagonalization, and aims to show that no language can
express all properties, since by diagonalization we can construct a
property that wasn’t expressible in that given language.21 This argu-
ment tries to show that the properties expressible in a given language
are a proper subset of all the properties, but it doesn’t try to do this
20 This is, of course, Grelling’s paradox.21 These arguments are called diagonal argument since they use the ‘‘diagonal’’ R (x, x)
of a binary relation R (x, y).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 187 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 187
using a cardinality consideration. There are many ways in which this
argument can be formulated more precisely. The details of the formu-
lation aren’t important for our present discussion, as we will see. I’ll
give an example of an argument using diagonalization that aims to show
that internalism has to be false. Here is one:
In every language there will be some property not expressible in that
language. Suppose you have some language l , and let’s assume that l
can talk about its own syntax, say via some coding. We’ll see that
there is a property not expressible in l . Consider a language l � which
extends l in the following way. l � has a two-place predicate SATl such
that SATl (x, y) holds just in case y is a code for the formula F of l and
F holds of x. Now, take the property expressed by :SATl (x, x), call it
D. D is not expressible in l . Suppose it is. Then there is some predicate
of l that expresses it, say C. It will have some code, say z. Then C (z)
holds iff :SATl (z, z) holds. But the latter holds just in case C(z)
doesn’t hold. Contradiction. So, :SATl (x, x) is not expressible in l .
The above argument doesn’t show what one might think it shows for
our discussion here. The language l indeed can’t express the property D
unless it gives rise to semantic paradoxes. If l already contains its own
satisfaction predicate then the extension to l � would lead to nothing
new. But then, of course, we could formulate a semantic paradox in l ,
using exactly the argument that was given above. So, the argument only
shows that there is a property that l doesn’t express under the assump-
tion that l does not contain its own satisfaction predicate. Now, the
natural language English does contain its own satisfaction predicate. It
can be expressed with the words ‘‘(the English expression) y truly
applies to x’’ or ‘‘(the English expression) y is satisfied by x’’. So, if
the above language l is supposed to be English then the extension to l �
is not a proper extension. l � is just English again. The argument given is
then simply a version of the paradox we discussed in the previous
section. The diagonal predicate :SATEnglish (x, x) is nothing but a formal
version of the predicate ‘‘does not apply to itself’’.
The argument thus does not show that there are properties inexpress-
ible in English. It only shows that if a language does not contain its own
satisfaction predicate then it doesn’t express all properties, and if it does
contain its own satisfaction predicate then it gives rise to paradoxes. One
can make the same point more easily by considering truth instead of
satisfaction. If a language does not contain its own truth predicate then
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 188 19.10.2005 3:15pm
188 | Thomas Hofweber
it does not express all properties (namely the property of being a true
sentence of that language), and if it does contain its own truth predicate
then it gives rise to paradoxes. For the natural language English the
latter of these options applies in both cases. In particular one should thus
not conclude from the above argument that there are limits to what can
be expressed in our own language.
How the paradoxes are to be dealt with is, again, another story, one
that the internalists and the externalists alike will have to deal with.
Diagonal arguments of the above kind point to the paradoxes, not to
limits in what is expressible.22
4.5. Modal Arguments
Somewhat different arguments against internalism uses modal claims of
some kind or other. Such arguments claim that internalism gets the
truth value of certain modal claims wrong. In this section I will look at
several examples of this and argue that they provide no problem for
internalism rightly understood. Here it will be important to distinguish
claims that we have good reason to believe to be true no matter what
philosophical theory about modality or the nature of properties we
adopt, and claims that themselves express a substantial metaphysical
view about modality or properties. The latter, of course, can’t always be
accommodated. Internalism is itself one of the possible views about talk
about properties, and it can’t accommodate claims that characterize
opposing views. However, there are no modal claims I know of that
seem to be clearly correct and that are incompatible with internalism.
Let’s look at some examples. A common strategy to argue against
internalism is to claim that internalism gets wrong the truth value of
utterances in counterfactual situations in which there are no language
users. The idea is that since internalism uses conjunctions and disjunc-
tions over classes of predicates (or sentences) that in such circumstances
the disjuncts would be empty. But this is, of course, mistaken. A
sentence such as
(45) If there had not been any languages then Fido would nonethe-
less have a property,
comes out as
22 I’m indebted to Rich Thomason for pushing the issues in this section.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 189 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 189
(46) If there had not been any languages then nonetheless VP (Fido
is P),
which is equivalent to
(47) If there had not been any languages then nonetheless either Fido
is a cat, or Fido is a dog, or Fido is . . .
which, of course, is true. The underlying mistake here is to think that
internalists equate quantification over properties with quantification
over predicates. They don’t. They merely make the claim that quantified
statements over properties are equivalent to infinite disjunctions and
conjunctions, all of which can be formulated in one’s language and are
sufficiently similar.
Another example apparently in conflict with internalism is
(48) There might have been different fundamental physical proper-
ties.
This is ambiguous between:
(49) Different properties might have been the fundamental physical
properties.
(50) There might have been different properties than there are, and
some of them might have been fundamental physical properties.
The first is no problem for internalism, since internalism can easily
grant that even though being an electron is not a fundamental physical
property, it might have been. So, this reading of (48) just says that
different properties than the ones that are actually the fundamental
ones might have been the fundamental ones. Accepting this as well as
rejecting it is compatible with internalism.
The second reading of (48) requires that there might have been
different properties than there are, and some of them might have been
fundamental. This consists of two conjuncts. The second one poses the
same problem as the first reading of (48) and thus is no problem. We will
have to look at the first conjunct:
(51) There might have been different properties than there are.
Acceptance of this is closely tied to acceptance of
(52) There might have been different objects than there are.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 190 19.10.2005 3:15pm
190 | Thomas Hofweber
In fact, there are some plausible considerations that (52) implies (51).
Here is a common argument for this. I’ll formulate it as it is usually
given first, in an externalist framework. After that we will see how it
carries over to an internalist framework.
Suppose you think that there might have been some objects that in
fact there aren’t. Let Joe be one of them. Then being Joe’s brother is an
object-dependent property. Object-dependent properties, just like ob-
ject-dependent, or singular, propositions, exist only if the objects on
which they depend exist. So, the property of being Joe’s brother exists
only if Joe exists. And if Joe might have existed, but doesn’t, then there
might have been a property which in fact there isn’t, namely being Joe’s
brother. Or so the externalist’s reasoning.23
Now, according to internalism a quite analogous situation obtains.
What properties there are is affected by what objects there are. This
comes from the interaction of the infinite disjunctions with the first-
order variables that are bound from the outside. In our case, we have
(53) It is possible that there is a property P such that there actually is
no property Q such that P ¼ Q.
According to internalism this is equivalent to
(54) }(9~yy _P [nomP] is a property & @:9~ww _Q [nomQ] is a property &
[nomP] ¼ [nomQ])
And this is true if there might have been different objects than there are.
If that is so then the variables~vv and ~ww can range over different domains
and thus there can be instances of P and Q that will be different. Just
take being identical to a and being identical to b, for a an object which
there might have been, but in fact there isn’t, and b any object which in
fact exists.24
Internalism does not have absurd consequences about the truth values
of modal claims. This is not to say that internalism will be able to
accommodate everyone’s intuitions about modal claims. Some modal
claims will be incompatible with internalism. But such claims will
themselves express substantial metaphysical views. Internalism is one
23 Not everyone accepts object-dependent properties, of course. This example is merelyused to show how it is often argued that what properties there are depends on what objectsthere are, and how internalism mirrors this reasoning.
24 To be precise, the above argument uses the principle that what objects are not namedin our language can differ from one world to another.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 191 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 191
contender among such views and thus shouldn’t be rejected because it
doesn’t agree with its competitors.
5. consequences and applications
5.1. Summary about Properties
We have started out by looking at two general views about talk about
properties. One view, externalism, takes such talk to be about some
mind-and language-independent domain of entities. The other view,
internalism, takes quantification over properties to be merely a logical
device to increase expressive power in a language-internal way. It claims
that such quantified statements are truth-conditionally equivalent to
infinite disjunctions and conjunctions over the instances that can be
formulated in that language. Internalism seems to have a serious prob-
lem with inexpressible properties that quite directly seems to refute it.
We have seen in the above sections that this is false. Internalism,
understood not as extreme internalism but as moderate internalism,
resists attempts to refute by such arguments. Moderate internalism can
accommodate the fact that there is a sense in which different languages
express different properties. Moderate internalism is compatible with
the inductive argument, and with several more technical arguments that
attempt to refute internalism with inexpressible properties. We have
seen no reason to reject moderate internalism from considerations about
inexpressible properties.
If what I have said so far is correct then internalism is a serious
contender in the debate about properties. Of course, we have seen no
reason so far why one might think that internalism is true. This is a
substantial and further question. To decide whether or not internalism
or externalism is true we will have to look at different issues. Among
them are
1. What is the function of our talk about properties? Why do we talk
about them in the first place?
2. What is the semantic function of property nominalizations, such
as ‘‘being a dog’’? What is the correct understanding of quantifier-
free talk about properties, such as ‘Being a philosopher is fun,’ or
‘Redness is a sign of ripeness’?
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 192 19.10.2005 3:15pm
192 | Thomas Hofweber
3. What is the correct understanding of ordinary uses of quantifiers
over properties?
These are the important and hard questions. Strict internalists, loose
internalists, and externalists will differ in the answers they will give to
these questions. And who is right in the end depends on who has the
better answers to these, and other, questions. A direct refutation of
internalism with expressibility considerations is not going to work.
Once that is clear we can focus on the important questions. In a series
of other papers I have given more positive reasons to accept a version of
strict internalism. This internalist view is based on a view about quan-
tification in natural language, the relation between quantifiers and
ontology, a defense of an internal–external distinction about ontological
questions, the semantic function of that-clauses and property nominal-
izations, and other issues.25
There are, however, some important questions about expressibility
that are closely related to the debate we have been engaged in so far.
These are about how different languages do in fact differ in what can be
expressed in them. But before we can look at this, let’s see how what has
been said so far applies to talk about propositions, and why this debate is
important more generally.
5.2. Inexpressible Propositions
So far we have focused on properties. The same considerations and
arguments, however, carry over to propositions. We will not go through
the same issues again in the case of propositions. This would not only be
tedious, but is also unnecessary, since one can argue directly that there
are inexpressible properties if and only if there are inexpressible pro-
positions. If f is an inexpressible property then
(55) that Joe is f
is an inexpressible proposition. And if that p is an inexpressible prop-
osition then
25 For the main idea of this view and some issues that motivate it, see Hofweber (2005,and forthcoming a) with an emphasis on noun phrases, or (2000) with an emphasis onquantifiers. For more on noun phrases from a more general point of view, see Hofweberand Pelletier (forthcoming). For a discussion about quantification and ontology, and theinternal – external distinction, see (forthcoming Hofweber, b or 2005).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 193 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 193
(56) believing that p
or
(57) being hungry even though p
is an inexpressible property.
But besides that, it seems clear that whatever reason one might have
to believe in inexpressible propositions will give rise to reasons to
believe in inexpressible properties and vice versa. These issues run in
parallel, and so do the arguments to the conclusion that they provide no
problem for internalism. The moderate internalist’s proposal about the
modeling of the truth-conditions of quantification over properties car-
ries over directly to a model for the truth-conditions of quantification
over propositions. The only difference is that properties are shadows of
predicates, whereas propositions are shadows of sentences. According to
internalism, quantification over propositions is equivalent to infinite
disjunctions and conjunctions, but now the instances of the disjunctions
and conjunctions involve sentences, not predicates. And just as in the
case of properties, the relevant disjunctions and conjunctions will in-
volve extra free variables that are bound from the outside by infinitary
quantifiers.
5.3. An Application: Minimalist Truth
The claim that there are inexpressible properties and propositions plays
an important role in the endgame of several philosophical debates. One
obvious case is the debate about the metaphysics of properties and
propositions itself. But this is not the only case. I would like to point
to one example here where internalism has a quite direct impact and
would substantially help a certain position in an important debate, the
debate about minimalist theories of truth. I will restrict the applications
of internalism to other debates to this example, and I will try to be brief.
The assumption that there are inexpressible propositions gives rise to
a powerful argument against minimalist (or deflationary) theories of
truth. Such theories claim that truth is really a metaphysically thin
notion whose function is mainly to give us increased expressive power
of certain sorts. In particular, minimalists about truth stress the import-
ance of the Tarski biconditionals
(TB) It’s true that p iff p.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 194 19.10.2005 3:15pm
194 | Thomas Hofweber
(or an appropriate generalization thereof) for an account of the function
of talk about truth. They closely tie the function of talk about truth to
the increased expressive power it gives rise to, and they claim that truth
is a metaphysically thin notion. Minimalism about truth is thus motiv-
ated by considerations congenial to the ones that motivate internalism.
One of the standard objections against minimalist theories of truth,
the incompleteness objection (see Schmitt 1995: 141), relies on there
being inexpressible propositions. The objection goes thus:
Since there are propositions inexpressible in present English the
concept of truth isn’t captured by all the instances of (TB). The
predicate ‘true’ applies even to propositions that can’t be expressed
in our language, as in
(58). What the best philosopher in the year 3000 will write will be
true.
What this philosopher will write might not be English, and might not
be translatable into English. But our concept of truth nonetheless
applies. So, the concept of truth goes further than minimalists would
have it, we need a substantial notion of truth that applies more widely
and more generally.
To accommodate ascriptions of truth to inexpressible propositions, or
sentences, or utterances, in foreign languages that can’t be translated
into English, minimalists have taken quite radical measures, and they
have thereby made their views quite implausible. Paul Horwich (1990),
for example, thinks that the axioms of the theory of our truth predicate
consist in all the propositions of the same form as the ones expressed by
the instances of (TB). Many of them will not be expressible in English
but nonetheless they are axioms of the theory of our concept of truth.
Hartry Field (1994), bases his form of minimalism not on ascription of
truth to propositions, but on ascription of truth to utterances. He also
accepts that there are utterances that express propositions that are not
expressible in our present languages, and bites the bullet by accepting
that our concept of truth can meaningfully be applied only to utterances
that one can understand. Truth of other utterances makes sense only via
some translation to ones that one can understand, and doesn’t make
sense to ones that can’t be translated.
The viability of internalism is obviously most central to this debate,
and in particular to what formulation a minimalist theory of truth
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 195 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 195
should take. Neither Horwich nor Field give any arguments for there
being inexpressible propositions. That there are is a shared assumption
in the debate. Moderate internalism can help a minimalist theory of
truth in dealing with this objection.
At the beginning of this chapter we discussed the sentence (1) involv-
ing a truth predicate where it gives rise to increased expressive power.
Sentence (1) involves quantification over propositions:
(59) For all p if Jones said that p in the trial then it’s true that p.
If internalism is correct about quantification over propositions then
truth is attributed only to expressible propositions (ones that are loosely
speaker expressible). Thus the incompleteness objection vanishes. The
same holds for (58):
(60) For all p if the best philosopher in the year 3000 writes that p
then it’s true that p.
That minimalists could deny that there are inexpressible propositions is
a well-known option, but this option is always portrayed as a last and
desperate move. For example, Schmitt (1995: 142) writes that ‘‘This way
of replying must surely be a last, heroic resort,’’ and later calls it ‘‘in the
realm of the preposterous’’ (ibid.). But with the distinction between
extreme and moderate internalism, and with keeping different notions
of expressibility apart, we have seen that this is not at all so.26
5.4. Isn’t There More to the World Than What We Can Say about It?
Even if there are no direct arguments against internalism using ex-
pressibility considerations, one might feel a bit of unease with the
internalist view about expressibility. According to an internalist there
is an important sense in which the world can be fully described, and
all objects can be completely characterized, by us. But we might ask
for an explanation how that could possibly be so. Wouldn’t it be a
complete miracle if we could say everything there is to say about the
world we live in? And even if we can say everything there is to say, why
is this so?
26 In what sense different languages differ in expressive strength will be discussedshortly.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 196 19.10.2005 3:15pm
196 | Thomas Hofweber
There is one clear way in which this can be made sense of, but this is
hardly a way that an internalist would want to be committed to. It would
be no wonder that we can express all properties that objects have if a form
of idealism is true. If the world and the objects in it are a product or
construction of our mind then it might be no wonder that our expressive
resources completely capture the world. An idealist could claim that the
world as we experience and describe it is a product or construction of our
minds, and that our concepts play a central role in this construction.
Without going into the details, it might seem possible that an idealist of
this kind could claim that all aspects of the world can be captured in our
language since after all the world is our product, and it should be no
surprise that our products can be fully captured by us.
The internalist’s explanation of why we can say everything there is to
say is different. It is not because the objects we describe somehow
depend on our descriptive abilities. Rather it is based on a view about
what it is that we ask when we ask about expressing everything. The
internalist’s explanation for why we can express everything is based on
a view about how general talk about properties and propositions, talk
that involves quantification over them, relates to particular talk about
properties and propositions, one that doesn’t involve quantifiers over
them, or that is not about properties and propositions at all. According
to the internalist quantification over properties and propositions is
merely a generalization over the instances, rightly understood, in
one’s own language. Thus when we ask whether or not we can say
everything there is to say we quantify over propositions. And if intern-
alism is right then this quantifier will be a generalization over the
instances in our own language. Thus no wonder the answer is that we
can say everything there is to say.
The argument that an explanation for the alleged fact that we can fully
describe the world requires a form of idealism or pure luck is based on an
externalist thinking about properties. True enough, for an externalist
complete expressibility would be a surprise and would require a substan-
tial explanation. If properties and propositions are out there independ-
ently of us, it would indeed be a surprise if we can express them all.
An internalist can and should claim that the world and its objects exist
independently of us and that objects have the properties they have
independently of us.27 Objects have properties independently of us in
27 Except, of course, response-dependent properties and the like.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 197 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 197
the sense discussed above, namely that Fido would still have the prop-
erty of being a dog even if there were no humans, which comes down to
that Fido would still be a dog even if there were no humans. This is the
beauty of internalism: the world and the objects in it exist independ-
ently of us, objects have the properties they have independently of us,
but still, properties are mere shadows of predicates, our predicates.
So far we have mainly focused on arguments that attempt to refute
internalism using considerations about expressibility. It is now time to
look at a more positive proposal about expressibility.
6. the expressibility hypothesis
6.1. The Hypothesis
Everyone agrees that in some sense different languages differ in their
expressive strength. What properties and propositions speakers of these
languages are able to express differs in some sense. The tricky part is to
say more precisely in what sense they differ and why and how this
difference comes about. And, of course, there is the additional tricky
question whether or not there is a sense in which they do not differ. In
the above we have seen two important ways in which different lan-
guages differ with respect to expressive strength. First, there can be a
difference between what can be expressed using eternal sentences or
predicates in different languages. Secondly, there is a difference be-
tween what contexts are in fact available to speakers of a certain lan-
guage. These two differences in expressibility are clearly present, and it
is not too hard to see why they are there (we will talk more about this
below). But are there any substantially different ways in which different
languages differ in expressive strength? And are there limits to what
languages in general can express? What we have seen so far suggests a
hypothesis about expressibility that says ‘no’ to these questions. It gives
rise to a picture about expressibility that is congenial to internalism and
to a picture about how and why languages differ in what can be said with
them:
(EH) The Expressibility Hypothesis. Different languages can differ in
what can be expressed in them with context-insensitive expressions,
and what speakers of these languages can in fact express in them.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 198 19.10.2005 3:15pm
198 | Thomas Hofweber
However, all languages agree on what speakers can express with them
in arbitrary contexts.28
Of course, not every system of symbols deserves to be called a ‘lan-
guage’. The expressibility hypothesis is clearly false if we allow traffic
signs, or Cþþ, or the dance of a been to be a language. These should
clearly not count as a language for our purposes, and all first human
languages should clearly count as languages. How to mark the differ-
ences more precisely is a substantial and interesting question that we
won’t be able to address properly here. The expressibility hypothesis is
interesting and controversial enough even when restricted only to
human languages. But as a first approximation, for a system of symbols
to be a language it has to satisfy at least some minimal conditions: it has
to allow for the expression of basic logical concepts, as well as certain
other basic concepts. These basic concepts will be discussed further
below. The expressibility hypothesis can be empirically refuted, and
empirically confirmed, by considering how different languages in fact
differ from each other in expressive strength. This is a substantial task,
and the considerations given below in support of the expressibility
hypothesis can only be considered to be a sketch of an outline of issues
that deserve a more thorough investigation. Nonetheless we shall have a
closer look at the expressibility hypothesis and how it relates to intern-
alism in this chapter.
6.2. Internalism and the Expressibility Hypothesis
The expressibility hypothesis and internalism about talk about proper-
ties and propositions are congenial, but independent. Internalism does
not imply the expressibility hypothesis. Internalism and expressive
chauvinism29 are consistent. Internalism might be true, and it might
28 This hypothesis has to be distinguished from one that has been endorsed by Searle(1969). Searle’s hypothesis is that
(H1) For every proposition p, if you can think that p then you can say that p.
i.e. the content of any thought can be articulated in language. The present expressibilityhypothesis is different. It states that
(H2) For every proposition p, if someone can say that p in some context theneveryone can say that p in the right context.
i.e. everyone can express any proposition that anyone can express, in the right context.29 See s. 3.1, where this notion was introduced.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 199 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 199
be true that our language expresses everything there is to express, even
though other languages don’t. Not that we should believe this. These
two positions are, however, consistent with each other.30 But the
expressibility hypothesis implies that expressive chauvinism is false. If
it is true then all languages will be equal when it comes to what speakers
can express with them in arbitrary contexts, which is what the crucial
clause of the expressibility hypothesis says:
(61) For every proposition p, if a speaker of some language l 1 can
express p in some context then for every language l 2 there is
some context C and some sentence F of l 2 such that an utterance
of F in C by a speaker of l 2 expresses p.
So, internalism about talk about properties and propositions can be true
and at the same time the expressibility hypothesis can be false. So, the
former does not imply the latter.
And the expressibility hypothesis does not imply internalism. It is
consistent with the expressibility hypothesis that propositions are
language-independent entities, and that there are propositions that are
not expressible in any language, in any context. Let’s call a proposition
that is not loosely speaker expressible in any language a completely alien
proposition. That there are completely alien propositions is consistent
with the expressibility hypothesis, but not with internalism. Sentence
(61) says only that what is loosely speaker expressible in one language is
loosely speaker expressible in any other. This is consistent with the claim
that some propositions are not loosely speaker expressible in any
language. Internalism requires, however, that all propositions are
loosely speaker expressible in our own language. Thus the expressibility
hypothesis does not imply internalism.
But internalism and the expressibility hypothesis go together. To
accept the expressibility hypothesis and deny internalism would be to
accept that even though any proposition expressible in a language at all
is expressible in every other language, there nonetheless are completely
alien propositions, propositions not expressible in any language what-
soever. It is hard to see what reason one might have for this view. And to
accept internalism but deny the expressibility hypothesis would require
30 It might even be that internalism is true for every language, but still our languageexpresses more than every other one. If internalism is true for some language l thenquantification over propositions in that language will be equivalent to infinite disjunctionsand conjunctions formulated in that language. It might nonetheless be so that onelanguage can express strictly more than another.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 200 19.10.2005 3:15pm
200 | Thomas Hofweber
one to accept a form of expressive chauvinism. Again, it isn’t easy to see
how this could be justified.
6.3. Expressive Change
To see that the expressibility hypothesis is a reasonable proposal, let’s
look a little bit more closely at how languages differ in expressive
strength, and why they differ in it in this way. It is particularly in-
structive to look at how the expressive power of a language differs over
time. One very important aspect of this is the process of de-contextual-
ization. In this process a language is modified in such a way that
speakers of it are able to express a certain process without them having
to be in a certain context. A good example of this is the introduction of a
name into the language.31 Suppose I like to talk about a certain object o.
I can always do that if o is around, and I can refer to o using a
demonstrative. But when o is not around and isn’t in the right spatial
or temporal relation to me then talk about o becomes impossible or at
least tedious. To get around the requirement of having to be in a special
spatial or temporal relationship with o I can introduce a name for o. The
expanded language I now speak relieves me of the requirement of being
related to o in a certain way to be able to talk about o. But what I can say
now (in any context) and what I could have said before (in special
contexts) is the same. The only thing that has changed is that the
expansion of the language has made what I can say more independent
of what contexts I have to be in to say it. In this chapter we have focused
mainly on reference to an object in a context. Other ways in which the
context of a speaker might have an impact on what can be said can be
understood analogously.
A second, somewhat more general, way in which languages evolve
over time is lexical addition. In this process the language gets expanded
with a new word that expresses what was before only expressible in a
complicated way. A good example of this is to introduce a new, simple
predicate for a complex phrase that for some reason or other has
acquired greater importance over time, and been used more and more
often. This would be analogous to an explicit definition, in the simplest
case. But this way can be and often will be mixed with the first way in
which languages differ. A new word will be introduced that allows
31 Not via a description. The case of introducing a name via a description is discussedbelow.
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 201 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 201
speakers of the expanded language to communicate in a simple way
independently of being required to be in a certain context that before
could only be said in a complicated way while being in a certain context.
A mixed case is introducing a name via a description. Not only does it
allow one to continue talking about an object even if it changes its
properties, it also allows one to talk about it in a simpler way than by
having to pick it out via a possibly quite complicated description. In
addition, such a description might contain context-sensitive elements,
and introducing a name rids the speaker of having to be in a particular
context to describe the object successfully.
What de-contextualization and lexical addition will occur in a lan-
guage over time will heavily depend on what the interests and needs of
the speakers in the language community are. Objects that are important
will get named, complex phrases that are or become important will be
the basis for lexical addition. In addition, what objects and contexts are
in fact available to speakers of that language will have a great impact,
too, for how the language will evolve.
So, we see that there are several completely unproblematic ways in
which what can be said in a language, and how it can be said in different
languages, can differ:
� different languages differ in what can be said by speakers of them
independently of the requirement of being in a certain context;
� different languages differ in how easy it is to say something, i.e.
how many words are needed to say it;
� speakers of different languages will in fact have different contexts
available to them to make utterances in.
These ways in which languages differ are unproblematic and together
sufficiently strong to give rise to the impression that what can be
expressed with a language is something quite independent of and ex-
ternal to the language. These considerations account for many extern-
alist intuitions, and why the expressibility hypothesis seems radical.
But moderate internalism accommodates all of them. Is there any
reason to believe that there is a difference in what can be said in
different languages that goes beyond these, or analogous, consider-
ations? How might it come about that different languages indeed do
express propositions that can’t be expressed in the other language, even
in arbitrary contexts?
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 202 19.10.2005 3:15pm
202 | Thomas Hofweber
6.4. Holism and the Expressibility Hypothesis
One way in which I can see this to be the case is holism. If holism is right
then, leaving subtleties aside, there aren’t two sufficiently different lan-
guages that can express the same proposition. Holism is a real alternative
to the expressibility hypothesis, but I won’t argue with it here. I personally
find little reason to believe holism to be true, but if you do then the
expressibility hypothesis is not for you. Holism and the expressibility
hypothesis describe the two extreme ends of the spectrum about how
languages differ in what can be expressed in them. Holism says, again
leaving subtleties aside, that nothing that can be said in one language can
be said in another one. The expressibility hypothesis says, also leaving
subtleties aside, that everything that can be said in one language can be
said in every other one. If you are like me and you find no reason to accept
holism then the expressibility hypothesis should be a serious candidate for
you. It makes sense of how languages differ in expressive strength and
lexical set-up. It makes sense of why we think that what can be said in one
language isn’t all that can be said (by taking recourse to the different
notions of expressibility). And it gives us an account of how and why
different languages differ in what can be said with them without collapsing
into a position that doesn’t allow for the same thing to be expressible in
different languages, even radically different languages spoken by speakers
in different locations and at different times.
We who reject holism should take the expressibility hypothesis as a
working hypothesis about what is expressible and how different lan-
guages differ in what can be expressed in them. The expressibility
hypothesis might be too naive, and might ultimately have to be rejected.
But if it fails we should see why precisely it isn’t enough, and what
gives rise to different expressive strength that isn’t already captured by
the cases discussed above. There might be some reason to reject the
expressibility hypothesis as stated, but this reason is only a modification
of one of the themes we have already discussed. But maybe there is a
reason to reject the expressibility hypothesis altogether which is not
based on a particular wording of the hypothesis. If we could see such a
reason we would have learned something very substantial and import-
ant about our languages or our minds.
One tempting line of reasoning to refute the expressibility hypothesis
is based on the observation that what is expressible in a language
according to the expressibility hypothesis depends on the basic concepts
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 203 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 203
that are articulated in the language. Not everything expressible in a
language can come from de-contextualization and lexical addition, or
related processes. Some expressive resources have to be basic. Now,
suppose it is true that if two languages articulate the same basic concepts
then what is loosely speaker expressible by predicates and sentences in
them is identical. Might there not be different languages that express
different basic concepts? And if so, wouldn’t what is loosely speaker
expressible in these languages be different? Isn’t that reason to give up
the expressibility hypothesis?
It is certainly conceivable that other creatures speak a language that
contains different concepts as the basic concepts. This is compatible with
the expressibility hypothesis. One such case would be that even though
for us negation and disjunction are the basic truth functional operators,
for them it’s negation and conjunction. Still, though, for both of us all
truth functions are expressible. What is required for this kind of argu-
ment to refute the expressibility hypothesis is that there are creatures
whose basic concepts are not even expressible by us. And this is, of
course, much more controversial, and in fact repeats one of the contro-
versies we had above. Can we really make sense of the idea that other
people or other creatures have some basic concepts that we can’t express
at all? If you like holism you might find this plausible, and if you tap
your externalist intuitions you might have little problem with it. But
independently of that I see little reason to accept it, and to give such
considerations an important status in our deliberations about what large-
scale view about expressibility and talk about properties and propositions
we should accept. The expressibility hypothesis makes sense of why and
how languages differ in their expressive power. Unless we find plausible
reasons to reject it I think we should work with it. It’s our best bet.
7. conclusion
The main aim of this chapter was to show that internalism is not refuted by
considerations about expressibility. Internalism seems to be committed to
expressive chauvinism, and this is a very implausible view. However, once
we distinguish extreme from moderate internalism, and once we distin-
guish several notions of expressibility, it becomes clear that issues about
expressibility and how they relate to internalism are not as simple as they
might at first seem. As we have seen in this chapter, moderate internalism
can accommodate claims to the effect that for every real number there is a
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 204 19.10.2005 3:15pm
204 | Thomas Hofweber
property of being larger than that number, that there might have been
different properties than there are, and that not every property is express-
ible in every language (‘‘expressible’’ rightly understood). In addition,
internalism is congenial to a view about expressibility, captured in the
expressibility hypothesis, which is a plausible hypothesis about expressive
change, at least for those among us who don’t believe in holism. Intern-
alism is not implied by this view about expressibility, since this view does
not rule out that there are completely alien propositions. But should we
believe that there are propositions inexpressible by any speaker in any
language in any context? The answer to this question will partly depend
on the answer to the question of what we do when we talk about proposi-
tions in the first place. In this debate we try to understand our own talk
about properties and propositions. What function this talk has will be
central to determining whether or not internalism or externalism is true.
What we do when we talk about and quantify over propositions will be
part of the story about whether or not the statement ‘‘there are completely
alien propositions’’ is true. If externalism about talk about propositions is
right then it might be true, if there indeed is a completely alien proposition
out there. But if quantification over propositions has a function more
congenial to the internalist then it will be false, whatever is out there in
the world. Part of what is at issue in answering this question is under-
standing our own language, in particular what we do when we talk about
properties and propositions.
To decide how this talk should be understood will involve a number of
further and substantial issues, ones that do not directly relate to con-
siderations about expressibility. These will be issues partly in the
philosophy of language, and partly in metaphysics. They will include
issues about the function of property nominalizations and quantifier-
free talk about properties, whether or not that-clauses are referring
expressions, issues about the role of properties in accounts of laws of
nature and causation, and many more. These are the issues where the
debate between externalism and internalism, and ultimately between
minimalist and substantial approaches to the metaphysics of properties
and propositions, will be settled. How this will go is, of course, a
completely different story. The goal of the present chapter is merely
to make sure that internalism and minimalism are not ruled out because
of considerations about expressibility.
University of North Carolina at Chapel Hill
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 205 19.10.2005 3:15pm
Inexpressible Properties, Propositions | 205
references
Armstrong, D. (1978). A Theory of Universals. (Cambridge: Cambridge
University Press).
—— (1989). Universals: An Opinionated Introduction. (Boulder, Colo.:
Westview).
Bach, K. (1997). ‘‘Do Belief Reports Report Beliefs?’’ Pacific Philosophical
Quarterly, 78: 215–41.
Barwise, J. (1975). Admissible Sets and Structures (Berlin: Springer Verlag).
Carnap, R. (1956). ‘‘Empiricism, Semantics, and Ontology’’, in Meaning and
Necessity (Chicago: University of Chicago Press), 2nd edn.
Field, H. (1994). ‘‘Deflationary Views of Meaning and Content’’, Mind 103:
249–85.
Hofweber, T. (2000). ‘‘Quantification and Non-existent Objects’’, in A. Ever-
ett and T. Hofweber (eds.), Empty Names, Fiction, and the Puzzles of Non-
Existence (Stanford, Calif.: CSLI).
—— (2005). ‘‘A Puzzle about Ontology’’, Nous, 39: 256–83.
—— (forthcoming a). ‘‘Innocent Statements and their Metaphysically
Loaded Counterparts’’.
—— (forthcoming b). ‘‘Neo-Carnapian Approaches to Ontology’’.
—— and Pelletier, J. (forthcoming), ‘‘Encuneral Noun Phrases’’.
Horwich, P. (1990). Truth (Oxford: Blackwell).
Keisler, H. J. (1971). Model Theory for Infinitary Logic (New York: North-
Holland).
King, J. (2002). ‘‘Designating Propositions’’, Philosophical Review 111/3.
Quine, W. V. O. (1980). ‘‘On What There Is’’, in From a Logical Point of View
(Cambridge, Mass.: Harvard University Press).
Schiffer, S. (1987). Remnants of Meaning (Cambridge, Mass.: MIT).
—— (1994). ‘‘A Paradox of Meaning’’, Nous, 28: 279–324.
—— (1996). ‘‘Language Created, Language Independent Entities’’, Philo-
sophical Topics 24: 149–67.
Schmitt, F. F. (1995). Truth: A Primer (Bouldes, Colo.: Westview).
Searle, J. R. (1969). Speech Acts (Cambridge: Cambridge University Press).
Yablo, S. (2000). ‘‘A Paradox of Existence’’, in A. Everett and T. Hofweber,
(eds.), Empty Names, Fiction and the Puzzles of Non-Existence (Stanford,
Calif.: CSLI).
Dean W. Zimmerman / Oxford Studies in Metaphysics - Volume 2 07-Zimmerman-chap07 Page Proof page 206 19.10.2005 3:15pm
206 | Thomas Hofweber